AN EMPIRICAL STUDY OF SCRUM SPRINT DURATION VARIABILITY AND ITS RELATIONSHIP WITH DELAY FREQUENCY IN OPEN-SOURCE AGILE PROJECTS (USING THE TAWOS DATASET)

Abstract

Research on delay risk in Agile projects tends to emphasise story points, velocity, and issue-level signals, with calendar duration of sprints rarely treated as a project-level variable in its own right. The present study addresses this gap. Using the open TAWOS dataset, we examine the actual calendar duration of Scrum sprints, its variability across projects, and the link between that variability and how often projects miss planned dates. After a four-step filtering procedure the working sample contains 4335 closed sprints from 36 open-source projects over the 2012–2020 period. For every sprint, planned and actual duration are compared. At the project level, the coefficient of variation of actual duration (CV) serves as a measure of calendar rhythm discipline. Within the sample, 58.02% of sprints fall within ±10% of the planned duration; 22.58% exceed the plan by more than 25%; 14.33% show minor

delays of 10–25%; and 5.07% finish ahead of schedule. Across the 31 projects with at least 30 sprints each, CV of actual duration is positively associated with the share of major-delay sprints (Spearman ρ = 0.609, p = 0.000275) and inversely with the share of on-time sprints (ρ = –0.538, p = 0.002). The Kruskal–Wallis test indicates that sprint duration distributions differ substantially across projects (H = 1048.70, p < 0.001). Tertile stratification by CV produces three groups - disciplined, moderate, chaotic - that also differ by delay share (H = 8.03, p = 0.018). The share of major-delay sprints declines monotonically over the observation window (ρ = –0.767, p = 0.016), which may indicate the gradual maturation of Agile practices in the open-source community. The findings are positioned as an empirical baseline for future predictive work on Scrum delay risk, not as a stand-alone prognostic model. The proposed three-tier typology may be used as a descriptive stratification frame for initial assessment of calendar discipline in IT-outsourcing projects.

Keywords: Scrum, duration variability, TAWOS, agile projects, statistical analysis, coefficient of variation, project typology, delay risks.